Field of the Invention
The present invention relates to a distributed processing system and a method of distributed processing for executing a sequence of processing steps in a distributed fashion by use of a plurality of processors, for effectively performing troubleshooting and for making it possible to identify replicated data belonging to the same processing.
In the U.S. Pat. No. 4,627,055, assigned to the same assignee as the present application, there has been described a distributed processing method for executing a sequence of processing steps in a distributed manner by use of a plurality of processors connected via transmission lines. In the patented system programs are distributed to and stored in respective processors to execute the respective portions of a sequence of processing steps and an initiation of a program of each processor is conducted when all data necessary to execute the program is completely received via a transmission line into the pertinent processor. Although this method makes it possible to achieve a sequence of processing steps through a distributed processing in the respective processors each effecting an execution of a portion of the processing steps without requiring the use of a control processor for controlling the overall system, there is not provided means for collecting the operation history of the system necessary at an occurrence of an abnormality to analyze (to conduct a troubleshooting) the cause of the abnormality and hence there arises a problem with respect to the maintainability of the system.
While some conventional operating systems (OS's) have a function to collect the operation history of the system in some cases, the objective of such data collection includes only the operation history in the processor in which the pertinent OS is loaded, which is insufficient to accomplish the troubleshooting in a distributed processing system for executing a sequence of processing steps in a distributed manner by use of a plurality of processors.
The Japanese Utility model Laid-open No. 57-175238 discloses a method in which when there exist a plurality of processors for executing the same processing, in place of the operation to select a truth value by use of a voter from the respective processing execution results of the processors to send only the truth value to the transmission network, the processing execution results are sent to the transmission network without any operations conducted thereon and each processor on the receiving side selects from the messages on the network the message belonging to the same processing operation for the pertinent processor so as to select a truth value in the processor on the receiving side according to a majority decision logic. According to this method, messages having the same content code (indicating the content of the data) are collected for a fixed period of time to count the number of messages thus collected and based on the result of the count, the truth value is judged according to the majority decision logic. The invention described above provides a flexible voting method which avoids the predecision to beforehand determine a processor from which an output is to be received into the voter.
In the prior art technology, there is used a method in which a message which is transmitted on a transmission medium includes a code indicating the content thereof and the messages for which the code matches a stored code are collected by a processor as messages belonging to the same process for a period of time from the collected messages according to a majority decision logic. In this method, however, since the messages belonging to the same processing are identified only by the code indicating the content of each of the messages, when a processor for executing the same process is initiated by a different trigger during a short period of time less than the period of time for collecting the messages, there arises a problem that an output message associated with the different trigger cannot be identified.